Shikai Geng is a PhD student in the RoMA Lab at UCL (joined 2025), working on the foundations of Vision–Language–Action (VLA) models. He develops tools, datasets, and training protocols that enhance the reliability, transparency, and extensibility of VLA systems for the community. Current projects include designing richer evaluation metrics and benchmarking suites that assess robot action quality beyond a single “success rate” capturing temporal competence, task fidelity, safety, and sample efficiency, and exploring federated learning to enable cross-institutional training while preserving data privacy. Broadly, his interests span multimodal perception, 3D scene understanding, and learning-augmented control, with an emphasis on reproducibility and open, end-to-end pipelines. Before his PhD, Shikai completed an MSc in Robotics and AI at UCL and a BSc in Informatics at the University of Edinburgh.
Robotics & Multimodal Autonomy (RoMA) Lab
at University College London